Noticed this about write_program_memory

[mskala]

I noticed a problem with WRITE_PROGRAM_MEMORY that I haven't seen on the search.

I have a way around it easily, but it appears that if the first time after reset that you call it you are using an address that is 2 bytes after a 64-byte boundary, occasionally the written value will be incorrect.

I changed (to what I was intending anyway) to first write on 64-byte boundary and have had no issues.

This is on 18F4520, with version 4.030, using some mods to the loader.c example. Found it because without a correct first byte on a bootloaded program, it will bomb. But it was strange that it works sometimes.